A methodology based on Particle image velocimetry for river ice velocity measurement

The in-field use of Particle image velocimetry (PIV), or Large-scale PIV, may be complicated by implementation issues related to the movement of the camera, the variations in luminosity and the distortion of the images due to the inclination of the shore-based camera. The correction for the distortion of the images, or orthorectification, is a critical issue that usually requires the georeferencing of control points located on the river surface. In drifting ice conditions, control points cannot be identified directly on the river surface, and water levels are often unstable. A robust, flexible and relatively inexpensive methodology based on Large-scale PIV for the measurement of river ice velocity is presented. Two orthorectification techniques are proposed, that do not require georeferencing of control points directly on the river surface. The proposed methodology also includes image enhancement, correction for camera movements and a segmentation technique that allows dynamical quantitative estimations of the ice cover extent. The methodology is tested on the St. Lawrence and Montmorency rivers, two rivers of very different sizes and ice conditions. The correction for camera movements proves to be quite reliable and robust to illumination variations and to the presence of shade patterns. The two orthorectification techniques allow to estimate pixel physical dimensions with 6% and < 1% precisions, respectively. Velocities obtained for the St. Lawrence case are compared to velocities simulated by a numerical model. There is strong agreement in velocity values in moderate ice coverage conditions.